Performance enhancement of cyclone deoiling for oily sludge based on BP neural network model and genetic algorithm

Performance enhancement in cyclone deoiling systems plays a critical role in improving oily sludge treatment efficiency and cutting energy consumption. This paper delved into the mechanism of cyclone deoiling, constructed Back Propagation (BP) neural networks to analyze the predictive performance of average shear rate and pressure drop, and achieved collaborative optimization of the cyclone structure by using the Nondominated Sorting Genetic Algorithm II (NSGA-II), aiming to balance deoiling efficiency and energy consumption. Through a combination of Computational Fluid Dynamics (CFD) simulations and experimental validation, the study systematically revealed the changes in particle dynamics and properties of oil-based mud (OBM) cuttings before and after optimization. The results demonstrated that the BP model outperformed the response surface model, Support Vector Machine (SVM), and Random Forest (RF) methods in predicting the average shear rate and pressure drop. The optimal cyclone structure corresponded to an average shear rate of 3111.23 s⁻¹ (an increase of 24.62 %) and a pressure drop of 992.54 Pa (an increase of 5.64 %), with prediction errors reduced to 0.80 % and 0.56 %, respectively. CFD simulations showed that the radial coupling centrifugal separation factor increased to 3.49 times that before optimization, and the pressure drop increased by 5.05 %. In the experiment, at an inlet velocity of 19 m/s, the oil content dropped to 0.49 %, the deoiling efficiency increased to 95.07 %, and the pressure drop increased by only 66.67 Pa, which was highly consistent with the predicted and simulated results. This study's intelligent optimization method provides an efficient, low-energy solution for oily sludge treatment, supporting sustainable oilfield development.